Download presentation
Presentation is loading. Please wait.
1
Classic AI Search Problems
Sliding tile puzzles 8 Puzzle (3 by 3 variation) Small number of 8!/2, about 1.8 *105 states 15 Puzzle (4 by 4 variation) Large number of 16!/2, about 1.0 *1013 states 24 Puzzle (5 by 5 variation) Huge number of 25!/2, about 7.8 *1025 states Rubik’s Cube (and variants) 3 by 3 by * 1019 states Navigation (Map searching)
2
Classic AI Search Problems
2 1 3 4 7 6 5 8 Invented by Sam Loyd in 1878 16!/2, about 1013 states Average number of 53 moves to solve Known diameter (maximum length of optimal path) of 87 Branching factor of 2.13
3
3*3*3 Rubik’s Cube Invented by Rubik in 1974 4.3 * 1019 states
Average number of 18 moves to solve Conjectured diameter of 20 Branching factor of 13.35
4
Navigation Arad to Bucharest start end
5
Representing Search Arad Zerind Sibiu Timisoara Oradea Fagaras Arad
Rimnicu Vilcea Arad Sibiu Bucharest
6
General (Generic) SearchAlgorithm
function general-search(problem, QUEUEING-FUNCTION) nodes = MAKE-QUEUE(MAKE-NODE(problem.INITIAL-STATE)) loop do if EMPTY(nodes) then return "failure" node = REMOVE-FRONT(nodes) if problem.GOAL-TEST(node.STATE) succeeds then return node nodes = QUEUEING-FUNCTION(nodes, EXPAND(node, problem.OPERATORS)) end A nice fact about this search algorithm is that we can use a single algorithm to do many kinds of search. The only difference is in how the nodes are placed in the queue.
7
Search Terminology Completeness Time complexity Space complexity
solution will be found, if it exists Time complexity number of nodes expanded Space complexity number of nodes in memory Optimality least cost solution will be found
8
Uninformed (blind) Search
Breadth first Uniform-cost Depth-first Depth-limited Iterative deepening Bidirectional
9
Breadth first QUEUING-FN:- successors added to end of queue (FIFO)
Arad Zerind Sibiu Timisoara Oradea Fagaras Rimnicu Vilcea Arad Arad Oradea Arad Lugoj
10
Properties of Breadth first
Complete ? Yes if branching factor (b) finite Time ? 1 + b + b2 + b3 +…+ bd = O(bd), so exponential Space ? O(bd), all nodes are in memory Optimal ? Yes (if cost = 1 per step), not in general
11
Properties of Breadth first cont.
Assuming b = 10, 1 node per ms and 100 bytes per node
12
Uniform-cost
13
Uniform-cost QUEUING-FN:- insert in order of increasing path cost Arad
Zerind Sibiu Timisoara 75 118 140 Oradea Fagaras Rimnicu Vilcea Arad 140 151 99 80 Arad Oradea 75 71 Arad Lugoj 118 111
14
Properties of Uniform-cost
Complete ? Yes if step cost >= epsilon Time ? Number of nodes with cost <= cost of optimal solution Space ? Optimal ?- Yes
15
Depth-first QUEUING-FN:- insert successors at front of queue (LIFO)
Arad Zerind Sibiu Timisoara Oradea
16
Properties of Depth-first
Complete ? No:- fails in infinite- depth spaces, spaces with loops complete in finite spaces Time ? O(bm), bad if m is larger than d Space ? O(bm), linear in space Optimal ?:- No
17
Depth-limited Choose a limit to depth first strategy
e.g 19 for the cities Works well if we know what the depth of the solution is Otherwise use Iterative deepening search (IDS)
18
Properties of depth limited
Complete ? Yes if limit, l >= depth of solution, d Time ? O(bl) Space ? Optimal ? No
19
Iterative deepening search (IDS)
function ITERATIVE-DEEPENING-SEARCH(): for depth = 0 to infinity do if DEPTH-LIMITED-SEARCH(depth) succeeds then return its result end return failure
20
Properties of IDS Complete ? Time ? Space ? Optimal ? Yes
(d + 1)b0 + db1 + (d - 1)b bd = O(bd) Space ? O(bd) Optimal ? Yes if step cost = 1
21
Comparisons
22
Summary Various uninformed search strategies
Iterative deepening is linear in space not much more time than others Use Bi-directional Iterative deepening were possible
23
Island Search Suppose that you happen to know that the optimal solution goes thru Rimnicy Vilcea…
24
Island Search Suppose that you happen to know that the optimal solution goes thru Rimnicy Vilcea… Rimnicy Vilcea
25
A* Search Uses evaluation function f = g + h g is a cost function
Total cost incurred so far from initial state Used by uniform cost search h is an admissible heuristic Guess of the remaining cost to goal state Used by greedy search Never overestimating makes h admissible
26
A* Our Heuristic
27
A* QUEUING-FN:- insert in order of f(n) = g(n) + h(n) Arad Zerind
Sibiu Timisoara g(Zerind) = 75 g(Timisoara) = 118 g(Sibiu) = 140 h(Zerind) = 374 h(Sibiu) = 253 g(Timisoara) = 329 f(Zerind) = f(Sibui) = …
28
Properties of A* Optimal and complete
Admissibility guarantees optimality of A* Becomes uniform cost search if h = 0 Reduces time bound from O(b d ) to O(b d - e) b is asymptotic branching factor of tree d is average value of depth of search e is expected value of the heuristic h Exponential memory usage of O(b d ) Same as BFS and uniform cost. But an iterative deepening version is possible … IDA*
29
IDA* Solves problem of A* memory usage Easier to implement than A*
Reduces usage from O(b d ) to O(bd ) Many more problems now possible Easier to implement than A* Don’t need to store previously visited nodes AI Search problem transformed Now problem of developing admissible heuristic Like The Price is Right, the closer a heuristic comes without going over, the better it is Heuristics with just slightly higher expected values can result in significant performance gains
30
A* “trick” Suppose you have two admissible heuristics…
But h1(n) > h2(n) You may as well forget h2(n) Sometimes h1(n) > h2(n) and sometimes h1(n) < h2(n) We can now define a better heuristic, h3 h3(n) = max( h1(n) , h2(n) )
31
What different does the heuristic make?
Suppose you have two admissible heuristics… h1(n) is h(n) = (same as uniform cost) h2(n) is misplaced tiles h3(n) is Manhattan distance
32
Effective Branching Factor
Search Cost Effective Branching Factor A*(h1) A*(h2) A*(h3) 2 10 6 2.45 1.79 4 112 13 12 2.87 1.48 1.45 680 20 18 2.73 1.34 1.30 8 6384 39 25 2.80 1.33 1.24 47127 93 2.79 1.38 1.22 364404 227 73 2.78 1.42 14 539 113 2.83 1.44 1.23 16 Big number 1301 211 1.25 Real big Num 3056 363 1.46 1.26
33
Game Search (Adversarial Search)
The study of games is called game theory A branch of economics We’ll consider special kinds of games Deterministic Two-player Zero-sum Perfect information
34
Games A zero-sum game means that the utility values at the end of the game total to 0 e.g. +1 for winning, -1 for losing, 0 for tie Some kinds of games Chess, checkers, tic-tac-toe, etc.
35
Problem Formulation Initial state Operators Terminal test
Initial board position, player to move Operators Returns list of (move, state) pairs, one per legal move Terminal test Determines when the game is over Utility function Numeric value for states E.g. Chess +1, -1, 0
36
Game Tree
37
Game Trees Each level labeled with player to move
Max if player wants to maximize utility Min if player wants to minimize utility Each level represents a ply Half a turn
38
Optimal Decisions MAX wants to maximize utility, but knows MIN is trying to prevent that MAX wants a strategy for maximizing utility assuming MIN will do best to minimize MAX’s utility Consider minimax value of each node Utility of node assuming players play optimally
39
Minimax Algorithm Calculate minimax value of each node recursively
Depth-first exploration of tree Game tree (aka minimax tree) Max node Min node
40
Example Max 5 Min 4 2 5 2 7 5 5 6 7 10 4 6 Utility
41
Minimax Algorithm Time Complexity? Space Complexity?
O(bm) Space Complexity? O(bm) or O(m) Is this practical? Chess, b=35, m=100 (50 moves per player) 3510010154 nodes to visit
42
Alpha-Beta Pruning Improvement on minimax algorithm
Effectively cut exponent in half Prune or cut out large parts of the tree Basic idea Once you know that a subtree is worse than another option, don’t waste time figuring out exactly how much worse
43
Alpha-Beta Pruning Example
3 3 2 3 30 a pruned 32 a pruned 3 5 2 3 5 53 b pruned 2 3 5 1 2 2 1 2 3 5
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.